基于协作表示和正则化最小二乘法的多姿态人脸识别方法

摘要

Since the existing face recognition method can′t deal with the posture change better,a multi-pose face recogni-tion method based on collaborative representation(CR)and regularized least square(RLS)method is proposed. The generalized elastic model(GEM)is used to construct a 3D model of the human face image in the image library. The features in three direc-tions(yaw,pitching and rolling)of the 3D face posture are extracted to construct a three-dimensional collaborative dictionary matrix(CDM). The RLS method and CR classification method are adopted to recognize and classify the human faces. The experi-ments were carried out with AR and video hunman face databases. The results show that the method can effectively solve the face recognition with posture variation,has the robustness for expression and illumination changes,and high real-time perfor-mance.%针对现有人脸识别方法不能很好地处理姿态变化的问题,提出一种基于协作表示和正则化最小二乘法的多姿态人脸识别方法.利用通用弹性模型(GEM)将图库中的人脸图像构建成为一个3D模型,在3D人脸姿态的三个方向(偏航、俯仰和翻滚)上提取特征,构建一个三维协作字典矩阵(CDM),利用正则化最小二乘法(RLS)和协作表示(CR)分类法对人脸进行识别分类.在AR和视频人脸数据库上进行实验,结果表明,该方法能够有效地解决具有姿态变化的人脸识别,同时对表情和光照变化也具有鲁棒性,且实时性高.